References & Citations
Statistics > Methodology
Title: General Bayesian $L^2$ calibration of mathematical models
(Submitted on 1 Mar 2021 (v1), revised 8 Apr 2022 (this version, v2), latest version 25 Aug 2022 (v3))
Abstract: A general Bayesian method for $L^2$ calibration of a mathematical model is presented. General Bayesian inference starts with the specification of a loss function. Then, the log-likelihood in Bayes' theorem is replaced by the negative loss. While the minimiser of the loss function is unchanged by, for example, multiplying the loss by a constant, the same is not true of the resulting general posterior distribution. To address this problem in the context of $L^2$ calibration of mathematical models, different automatic scalings of the general Bayesian posterior are proposed. These are based on equating asymptotic properties of the general Bayesian posterior and the minimiser of the $L^2$ loss. The approach is validated and compared to traditional Bayesian calibration on a range of synthetic and real applications.
Submission history
From: Antony Overstall [view email][v1] Mon, 1 Mar 2021 17:06:43 GMT (51kb,D)
[v2] Fri, 8 Apr 2022 13:43:08 GMT (221kb,D)
[v3] Thu, 25 Aug 2022 12:48:13 GMT (267kb,D)
Link back to: arXiv, form interface, contact.